From b4988c18c46a61a041db7d4d019f151beead2fce Mon Sep 17 00:00:00 2001
From: Eduardo Lopez <5743857+loama@users.noreply.github.com>
Date: Mon, 17 Aug 2026 12:33:17 +0200
Subject: [PATCH] docs: fix language errors

---
 docs/models.tft.html.md                     | 2 +-
 nbs/docs/capabilities/predictInsample.ipynb | 2 +-
 nbs/docs/tutorials/large_datasets.ipynb     | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/docs/models.tft.html.md b/docs/models.tft.html.md
index 70ac00bb6..70c2604e9 100644
--- a/docs/models.tft.html.md
+++ b/docs/models.tft.html.md
@@ -453,7 +453,7 @@ To avoid information bottlenecks from the classic Seq2Seq architecture,
 TFT incorporates a decoder-encoder attention mechanism inherited
 transformer architectures ([Li et. al
 2019](https://arxiv.org/abs/1907.00235), [Vaswani et. al
-2017](https://arxiv.org/abs/1706.03762)). It transform the the outputs
+2017](https://arxiv.org/abs/1706.03762)). It transforms the outputs
 of the LSTM encoded temporal features, and helps the decoder better
 capture long-term relationships.
 
diff --git a/nbs/docs/capabilities/predictInsample.ipynb b/nbs/docs/capabilities/predictInsample.ipynb
index d23e3d5ae..22632cb87 100644
--- a/nbs/docs/capabilities/predictInsample.ipynb
+++ b/nbs/docs/capabilities/predictInsample.ipynb
@@ -20,7 +20,7 @@
     "\n",
     "*Use Cases*: \n",
     "* Debugging: producing insample predictions is useful for debugging purposes. For example, to check if the model is able to fit the train set.\n",
-    "* Training convergence: check if the the model has converged.\n",
+    "* Training convergence: check if the model has converged.\n",
     "* Anomaly detection: insample predictions can be used to detect anomalous behavior in the train set (e.g. outliers). (Note: if a model is too flexible it might be able to perfectly forecast outliers)"
    ]
   },
diff --git a/nbs/docs/tutorials/large_datasets.ipynb b/nbs/docs/tutorials/large_datasets.ipynb
index 437df57b0..43d24692d 100644
--- a/nbs/docs/tutorials/large_datasets.ipynb
+++ b/nbs/docs/tutorials/large_datasets.ipynb
@@ -14,7 +14,7 @@
    "source": [
     "The standard DataLoader class used by NeuralForecast expects the dataset to be represented by a single DataFrame, which is entirely loaded into memory when fitting the model. However, when the dataset is too large for this, we can instead use the custom large-scale DataLoader. This custom loader assumes that each timeseries is split across a collection of Parquet files, and ensure that only one batch is ever loaded into memory at a given time.\n",
     "\n",
-    "In this notebook, we will demonstrate the expected format of these files, how to train the model and and how to perform inference using this large-scale DataLoader."
+    "In this notebook, we will demonstrate the expected format of these files, how to train the model and how to perform inference using this large-scale DataLoader."
    ]
   },
   {